Empirical Studies of the Distribution and Feedback Mechanisms of Mobile App Stores
MetadataShow full item record
Mobile app stores are online stores where users can purchase and download mobile apps. The marketplace for apps has exploded over recent years with hundreds of thousands of apps, millions of dollars in revenue and millions of users. Mobile app stores offer unique distribution and feedback mechanisms. The distribution mechanism of app stores acts as a central repository for all apps with an option to automatically update apps on user's devices. The feedback mechanism of app stores is a central rating and review system for all apps in a store. Consequently, there is a need for research to explore such unique mechanisms. In this thesis, we analyze over 10,000 of the top free apps across all the app categories in the Google Play Store. We first examine the distribution mechanisms of app stores and its impact on the release speed of apps. We then study the feedback mechanism of app stores to observe how it allows for a large influx of reviews and the benefits of responding to reviews. Finally, we propose an approach to automatically label reviews to help cope with the large influx of reviews by summarizing them. In regards to the unique distribution mechanism, we observe that a subset of apps release very frequently - more than one release per week. Almost half of these apps do not provide users with a rationale for a new release. Developers are leveraging the distribution mechanism by choosing to rapidly release new versions of their app. In regards to the unique feedback mechanism, we find that most apps do not receive many reviews, however, approximately 1% of apps receive over 500 reviews per day. We find that responding to reviews can lead to a positive increase in the rating of the app. The feedback mechanism enables users to give rapid and rich feedback. Finally, we propose an approach to label user reviews automatically using 14 different labels that capture the user's perceived quality of an app. We demonstrate the usefulness of our labelling approach through three different analytics use case scenarios: competitive analysis, app store overview and anomaly detection.